pushpam14 commited on
Commit
ecbbd26
·
verified ·
1 Parent(s): 9c794e4

Add trained_scores.json for judges to inspect from Space repo root

Browse files
Files changed (1) hide show
  1. trained_scores.json +183 -0
trained_scores.json ADDED
@@ -0,0 +1,183 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "pushpam14/api-contract-validator-grpo-7b",
3
+ "benchmark": "api_contract_validator",
4
+ "scores": {
5
+ "find_type_mismatches": 0.75,
6
+ "validate_nested_objects": 0.5714,
7
+ "detect_breaking_changes": 0.6667,
8
+ "validate_response_schema": 0.3,
9
+ "validate_cross_field_constraints": 0.2857,
10
+ "validate_auth_request": 0.3333,
11
+ "trace_downstream_blast_radius": 0.99,
12
+ "propose_backward_compat_fix": 0.99,
13
+ "multi_service_cascade_fix": 0.99
14
+ },
15
+ "details": [
16
+ {
17
+ "task": "find_type_mismatches",
18
+ "score": 0.75,
19
+ "steps": 10,
20
+ "success": false,
21
+ "rewards": [
22
+ 1.0,
23
+ -0.1,
24
+ 1.0,
25
+ -0.1,
26
+ -0.1,
27
+ 1.0,
28
+ -0.1,
29
+ -0.1,
30
+ -0.1,
31
+ -0.1
32
+ ],
33
+ "computed_from": "rewards"
34
+ },
35
+ {
36
+ "task": "validate_nested_objects",
37
+ "score": 0.5714,
38
+ "steps": 15,
39
+ "success": false,
40
+ "rewards": [
41
+ 1.0,
42
+ 1.0,
43
+ 1.0,
44
+ -0.3,
45
+ -0.3,
46
+ -0.3,
47
+ -0.1,
48
+ -0.1,
49
+ -0.1,
50
+ 1.0,
51
+ -0.3,
52
+ -0.3,
53
+ -0.1,
54
+ -0.1,
55
+ -0.1
56
+ ],
57
+ "computed_from": "rewards"
58
+ },
59
+ {
60
+ "task": "detect_breaking_changes",
61
+ "score": 0.6667,
62
+ "steps": 20,
63
+ "success": false,
64
+ "rewards": [
65
+ 1.0,
66
+ -0.3,
67
+ 1.0,
68
+ 1.0,
69
+ 1.0,
70
+ 1.0,
71
+ -0.1,
72
+ -0.1,
73
+ -0.1,
74
+ -0.1,
75
+ -0.1,
76
+ -0.1,
77
+ 1.0,
78
+ -0.1,
79
+ -0.1,
80
+ -0.1,
81
+ -0.3,
82
+ -0.3,
83
+ -0.1,
84
+ -0.1
85
+ ],
86
+ "computed_from": "rewards"
87
+ },
88
+ {
89
+ "task": "validate_response_schema",
90
+ "score": 0.3,
91
+ "steps": 4,
92
+ "success": true,
93
+ "rewards": [
94
+ 1.0,
95
+ 1.0,
96
+ 1.0,
97
+ 0.15
98
+ ],
99
+ "computed_from": "rewards"
100
+ },
101
+ {
102
+ "task": "validate_cross_field_constraints",
103
+ "score": 0.2857,
104
+ "steps": 18,
105
+ "success": false,
106
+ "rewards": [
107
+ 1.0,
108
+ 1.0,
109
+ -0.3,
110
+ -0.3,
111
+ -0.3,
112
+ -0.3,
113
+ -0.1,
114
+ -0.1,
115
+ -0.1,
116
+ -0.3,
117
+ -0.3,
118
+ -0.3,
119
+ -0.3,
120
+ -0.1,
121
+ -0.1,
122
+ -0.1,
123
+ -0.1,
124
+ -0.1
125
+ ],
126
+ "computed_from": "rewards"
127
+ },
128
+ {
129
+ "task": "validate_auth_request",
130
+ "score": 0.3333,
131
+ "steps": 14,
132
+ "success": false,
133
+ "rewards": [
134
+ 1.0,
135
+ 1.0,
136
+ -0.1,
137
+ -0.1,
138
+ -0.1,
139
+ -0.1,
140
+ -0.1,
141
+ -0.1,
142
+ -0.1,
143
+ -0.1,
144
+ -0.1,
145
+ -0.1,
146
+ -0.1,
147
+ -0.1
148
+ ],
149
+ "computed_from": "rewards"
150
+ },
151
+ {
152
+ "task": "trace_downstream_blast_radius",
153
+ "score": 0.99,
154
+ "steps": 1,
155
+ "success": true,
156
+ "rewards": [
157
+ 1.6
158
+ ],
159
+ "computed_from": "rewards"
160
+ },
161
+ {
162
+ "task": "propose_backward_compat_fix",
163
+ "score": 0.99,
164
+ "steps": 1,
165
+ "success": true,
166
+ "rewards": [
167
+ 2.0
168
+ ],
169
+ "computed_from": "rewards"
170
+ },
171
+ {
172
+ "task": "multi_service_cascade_fix",
173
+ "score": 0.99,
174
+ "steps": 2,
175
+ "success": true,
176
+ "rewards": [
177
+ 1.6,
178
+ 2.0
179
+ ],
180
+ "computed_from": "rewards"
181
+ }
182
+ ]
183
+ }